Argentina’s national statistics agency, INDEC, has recently published its poverty statistics for the second semester of 2020, and they are not pretty. 42% of the country is under the poverty line, representing 31.6% of households; for the Buenos Aires Metro Area (the largest population center in the country, home to 14 million people) it is 44.3%, and it reaches 51% outside the city’s boundaries (with 10 million inhabitants). This is the highest figure in over 11 years, and it gets starker: child poverty has reached 57.7%, with 15.7% of those in extreme poverty.
Besides the other methodological issues, an interesting question to tackle would be why poverty is measured by semester and not by quarter (like, say, unemployment). The reason is that, besides simpler data collection, workers receive two bonus payments worth 50% of their last salary in June and December (known colloquially as the “aguinaldo”) and the data would be very erratic otherwise, as poverty would decrease in the first and third quarters (because of the aditional income) even under ceteris paribus.
What is poverty, and how we measure it
How to measure poverty is probably one of the most hotly debated questions in economics. Generally speaking, there are two definitions widely in place: absolute poverty, which measures the irreducible group that does not fulfill certain basic qualifications, and relative poverty, which defines a set of needs to be met based on social context.
Generally speaking, there are three main kinds of poverty indicators one might encounter. The first are the simplest - an income level below which you are considered poor; normally this is either an “arbitrary” amount (such as the 2 dollars per day the World Bank uses), or a percentage of the median income (this is a common measurement in Europe), and neither are common in Argentina.
The second type closely matches the absolute poverty definiion - unmet basic needs (NBI by its Spanish initials), which creates a list of requisites to be satisisifed ex-post. NBI poverty is measured by INDEC during the national census, every ten years, and requires a household meeting one of five requirements: crowded home, unsanitary housing, lack of access to sewage, minors not attending primary education, and more than four people per employed person and/or head of household with less than three years of compulsory education. This poses two obvious issues: the bar to clear for poverty is very fuzzy (in this case, it is obviously too low), and the indicator is very sensitive to how many requirements are needed and how many there are in total. Since the questionnaire has only been asked in 4 censuses (1981, 1991, 2001, and 2010), there isn’t much to say: NBI poverty has continuously declined, from 22.3% in 1980 to 9.1% in 2010.
The third measurement is the poverty line, which defines a minimum required income to (not) be poor. Consumption would be the clearest indicator for a standard of living, but that kind of data is harder to compile, so income can be used as a proxy. Argentina does this by building a basket for nutritional requirements (The Basic Food Basket or CBA): they calculate how many nutrients people need to survive, and how these needs are met; adjust for income level, education, area of residence, etc. This is the basis for the extreme poverty line, which only measures people who can’t afford to feed themselves. The poverty line includes a more complex basket of goods and services, plus people move around the life cycle (parents of young children and the elderly don’t have the same spending patterns). Since people tend to spend fixed fractions of their income on food, you can approximate the Total Food Basket (or CBT) by using that fraction, called Engel’s Coefficient, and multiplying the CBA by the inverse of Engel’s Coefficient (for example, Argentina’s EC is about 40%, so the inverse is 2.5).
Argentina has used a notoriously high definition for both types of poverty since a methodological change in 2016, which also means that comparisons between the present and historical periods are not especially easy. In the following graph, values along the same colored line are comparable, values between lines are not - except for the the yellow, red, and purple lines, which use the same methodology (they correspond to backwards-looking estimates using the current methodology).
The Poverty Wars, part 1
In the graph above, there are three distinct official methodologies: 1988-2003, 2003-2013 (green line), and the one that has been in use since 2016 (purple line). What is going on there?
The first difference, and the easiest to explain, is “blue to green”. The blue line only measured a specific month, and used a certain set of food basket definitions; the green line measured each semester. The green line can be split in two: 2003-2007, and 2007-2013.
The reason for this is simple: in 2007 then President (and current Vice President) Cristina Fernández de Kirchner staged an intervention against the official statistics agency. This happened due to a high profile confrontation between the Commerce Secretary, Guillermo Moreno, and the technical staff. Moreno demanded a list of companies that had raised their prices due to a surge in inflation, which pointed to violations of price control agreements - but humoring his request would have broken the law. After this demand was refused, Moreno and Kirchner had the organism’s upper ranks purged and replaced with more amenable professionals, who would ensure inflation figures matched the government’s expectations. This resulted in laughably low food basket estimates, which led to a dropping poverty rate - by 2013, the results were clearly unbelievable, since it pointed to the country having “less poverty than Germany” (a common critique of the government at the time), and the poverty indicator was discontinued altoghether in 2015.
When center-right opponent Mauricio Macri replaced Kirchner as head of state in 2015, his administration swiftly declared a “statistical emergency” and had INDEC’s leading staff replaced by widely respected economists. This resulted, by the end of the following year, in newly reliable indicators - most importantly inflation and poverty, but due to methodology changes that rendered them non-comparable to prior estimates. There have been numerous attempts to bridge the gap - most notably a series of works by Leopoldo Tornarolli that estimate the poverty rate using various techniques (links provided at the end of the post).
The Poverty Wars, part 2
A relevant question is who people trusted for poverty statistics during the “Dark Age”. The answer: the Catholic University of Argentina’s Observatory For Social Debt (UCA - ODSA). They use a different methodology from INDEC: rather than having either an NBI indicator or a poverty line, they use both.
It should be noted UCA uses its own sample with its own methodology, so these results ARE NOT comparable. The way it works is very simple: they ask questions for five NBI-type indicators, and measure income. This means that they create a multidimensional measurement of poverty - one with a presence of income and necessities simultaneously. They also use different NBI benchmarks: food (not having experienced hunger or malnutrition), sanitation (sewage and running water), housing (basically the same), educational achievement (complete education for all children and adults), and employment (one registered position for every four members of the household). Thus, poverty can be defined thusly:
What’s the controversy? In late 2019, UCA published a summary of their poverty report - with an explosive revelation: the percentage of the population with unmet needs exceeded 40% for the first time ever. The part that’s controversial isn’t the big headcount, but rather a series of highly unusual circumstances: rather than publishing their report, they posted an online summary that mentioned their methodology had been modified - and it included no clear comparison to their (already published) figures from previous years.
This kicked off a media debate, especially since after catching flak for their lack of methodological transparency, UCA’s research center went ahead and posted on Twitter that they nobody was supposed to think their figure was comparable to the official one (which nobody in the media treated as such), and that they had made their own estimates of the actually comparable number, which was between 32% and 36%.
Conclusion
Why is this relevant? For once, this proves how vital a reliable set of national statistics is. Without it, there’s not even a possibility of a consensus for solving social problems - the solution is simply to bury your head in the sand.
This also points to another issue: how a lack of clarity to an audience of non-specialists can result in straightforward data being interpreted in bewildering and partisan reasons. The sloppiness and ineffectiveness of UCA’s communications team led to their very worrying results being dismissed out of hand by half the country due to perceived partisanship from their source. While there are nuanced methodological discussions to be had, it is clear that not making your assumptions and definitions clear to “normal” readers is only a source of trouble.
Sources
INDEC, “Poverty In The Second Semester of 2020” - in Spanish
This Twitter thread by Leopoldo Tornarolli - in Spanish
Roser and Ortiz-Espina “Global Extreme Poverty - Evolution of Extreme Poverty By Country” from Our World In Data
INDEC, “The Measurement Of Poverty In Argentina” (2016) - in Spanish
Tornarolli, Leopoldo “The Evolution Of Poverty In Argentina In The Last 15 Years” (2018) - in Spanish
Tornarolli, Leopoldo “Comparable Series Of Poverty And Extreme Poverty: A Proposal” (2018) - in Spanish
Gonzalez Rozada, Martin “Historic Evolution Of Poverty And Income Inequality In Argentina” (2019) - in Spanish
Gonzalez Rozada, Martin “Incidence of Poverty in Argentina 1988-2017” - in Spanish
Tornarolli, Leopoldo “Exploring Changes in Poverty in Argentina” (2019) - in Spanish
UCA, “Poverty As Denial Of More Than Income” (2020) - in Spanish
UCA, “Poverty Beyond Income: Report On Multidimensional Poverty” (2020) - in Spanish